I responded earlier to this question, for authorities. Following is the explanation of why we intend to remove class bf:Annotation in BIBFRAME 2.0.
The BIBFRAME annotation model developed for BIBFRAME 1.0 is well over two years old. Our current thinking is that it is over-engineered, has some significant flaws, and is completely obsolete. As I was the principal editor I take responsibility for the engineering and flaws. At the time, I was just beginning to learn about annotations. Community development of an annotation model was far less mature, and, we believed, did not seem to address bibliographic requirements.
A lot has changed since then, most notably the chartering of a formal W3C Web Annotation Working Group. Prior to that there was simply an informal Open Annotation Community Group. I have been an active participant of the new W3C group and am confident that the annotation model will be general enough to handle any bibliographic requirements, without the need for any BIBFRAME specific properties of classes.
Web Annotations will allow anyone to annotate any Web resource, including any BIBFRAME resource. There will be nothing in BIBFRAME 2.0 to preclude that, and most likely nothing necessary to support it.
It may be useful to develop a Web Annotation profile for bibliographic information. It would be premature to begin that effort now because the Web Annotation Specifications are not yet stable, and also because we don’t really have a complete understanding of the requirements for bibliographic annotations. What we had thought were the basic annotation types for BIBFRAME don’t seem to be as useful or appropriate as we had thought, for example, cover art and table of contents, and most importantly, holdings, with the addition of Item as a new core class, and its decoupling from holdings.